Event
Proposes HuTact tactile-aware hand representation pretraining framework
1 reports · 1 sources
Yeonguk Yu et al. propose the HuTact framework, which uses large-scale human-object interaction (HOI) data to pretrain tactile-aware hand representations; the pretraining stage does not require real tactile signals, and the framework is then adapted using a small amount of unlabeled robot interaction data. The paper claims that after training on HOI pretraining plus unlabeled robot data, on the grasp prediction task it achieves 88.26% Macro-F1 using only 25% of labeled downstream data, comparable to 88.23% from training from scratch with all labeled data.
Report timeline
- HuTact: Pretraining Tactile-Aware Hand Representations from Human Hand-Object Interaction Data
The paper proposes the HuTact framework, which uses large-scale human hand-object interaction (HOI) data to pretrain tactile-aware hand representations. The pretraining stage requires no real tactile signals, and the model is then adapted with a small amount of unlabeled robot interaction data.
arXiv cs.RO ·